One Retrieval to Cover Them All: Co-occurrence-Aware Knowledge Base Reorganization for Session-Level RAG

2026-06-30

AI Summary

This work addresses the limitation of traditional retrieval-augmented generation (RAG) systems, which are optimized for single-turn queries and struggle to fulfill coherent information needs across semantic regions of enterprise knowledge bases in multi-turn conversations. The authors propose a conversation-aware retrieval approach that reorganizes the knowledge base via offline co-occurrence-aware clustering and dynamically expands cluster neighborhoods at query time to encompass the full informational scope of an ongoing dialogue. Innovatively, they replace per-query recall with conversation-level coverage as the core evaluation metric, enabling simultaneous knowledge base compression and efficient retrieval. Evaluated on the WixQA dataset, their method improves single-retrieval conversation coverage by 17% (reaching 58%), reduces the number of retrievals needed to achieve 70% coverage by 34%, and compresses the knowledge base to 20% of its original size, with consistent gains across diverse embedding models and functional domains.

Abstract

RAG systems retrieve documents optimized for answering one query at a time. Yet enterprise users arrive with sessions, that is, coherent episodes of related questions that span semantically distant parts of the knowledge base. We show that a single retrieval call over a standard knowledge base covers only 41% of a user's session-level information need. To close this gap, we reorganize the KB offline using co-occurrence-aware clustering and expand retrieval candidates through cluster neighborhoods at query time. On WixQA (6,221 enterprise support articles), our method raises single-query session coverage to 58% (+17% absolute; 95% CI: [14.1, 20.4]), reduces retrieval calls to 70% coverage by 34%, and compresses the KB to 20% of its original size, all consistently across four embedding models and six functional domains. We argue that session-level coverage, not single-query recall, should be the primary metric for enterprise RAG evaluation.

Problem

• session-level RAG
• knowledge base reorganization
• co-occurrence-aware retrieval
• enterprise information need
• multi-query coverage

Innovation

• session-level RAG
• co-occurrence-aware clustering
• knowledge base reorganization
• retrieval coverage
• enterprise QA